Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/3001#discussion_r112122018
--- Diff:
flink-connectors/flink-connector-kinesis/src/main/java/org/apache/flink/streaming/connectors/kinesis/FlinkKinesisConsumer.java
---
@@ -267,38 +293,84 @@ public void close() throws Exception {
//
------------------------------------------------------------------------
@Override
- public HashMap<KinesisStreamShard, SequenceNumber> snapshotState(long
checkpointId, long checkpointTimestamp) throws Exception {
+ public void snapshotState(FunctionSnapshotContext context) throws
Exception {
if (lastStateSnapshot == null) {
LOG.debug("snapshotState() requested on not yet opened
source; returning null.");
- return null;
- }
-
- if (fetcher == null) {
+ } else if (fetcher == null) {
LOG.debug("snapshotState() requested on not yet running
source; returning null.");
- return null;
- }
-
- if (!running) {
+ } else if (!running) {
LOG.debug("snapshotState() called on closed source;
returning null.");
- return null;
+ } else {
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Snapshotting state ...");
+ }
+
+ offsetsStateForCheckpoint.clear();
+ lastStateSnapshot = fetcher.snapshotState();
+
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Snapshotted state, last processed
sequence numbers: {}, checkpoint id: {}, timestamp: {}",
+ lastStateSnapshot.toString(),
context.getCheckpointId(), context.getCheckpointTimestamp());
+ }
+
+ for (Map.Entry<KinesisStreamShard, SequenceNumber>
entry : lastStateSnapshot.entrySet()) {
+
offsetsStateForCheckpoint.add(Tuple2.of(entry.getKey(), entry.getValue()));
+ }
}
+ }
- if (LOG.isDebugEnabled()) {
- LOG.debug("Snapshotting state ...");
+ @Override
+ public void initializeState(FunctionInitializationContext context)
throws Exception {
+ TypeInformation<Tuple2<KinesisStreamShard, SequenceNumber>>
tuple = new TupleTypeInfo<>(
+ TypeInformation.of(KinesisStreamShard.class),
+ TypeInformation.of(SequenceNumber.class)
+ );
+
+ offsetsStateForCheckpoint =
context.getOperatorStateStore().getUnionListState(
+ new
ListStateDescriptor<>(DefaultOperatorStateBackend.DEFAULT_OPERATOR_STATE_NAME,
tuple));
+
+ if (context.isRestored()) {
+ if (sequenceNumsToRestore == null) {
+ sequenceNumsToRestore = new HashMap<>();
+ for (Tuple2<KinesisStreamShard, SequenceNumber>
kinesisOffset : offsetsStateForCheckpoint.get()) {
+
sequenceNumsToRestore.put(kinesisOffset.f0, kinesisOffset.f1);
+ }
+
+ LOG.info("Setting restore state in the
FlinkKinesisConsumer.");
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Using the following offsets:
{}", sequenceNumsToRestore);
+ }
--- End diff --
I would simply include `sequenceNumsToRestore` as part of the info log
message.
It's actually quite useful information for debugging sometimes, and users
normally do not enable debug level for the production jobs.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
